'HWHUPLQDQWVRIWKHSHUFHQWDJHRISUHIHUHQWLDOYRWHV DQGWKHHIIHFWRIHPSRZHULQJLQGLYLGXDOYRWHUV 7KHFDVHRIWKHORFDOHOHFWLRQVLQ)ODQGHUV%HOJLXP

'HWHUPLQDQWVRIWKHSHUFHQWDJHRISUHIHUHQWLDOYRWHV DQGWKHHIIHFWRIHPSRZHULQJLQGLYLGXDOYRWHUV 7KHFDVHRIWKHORFDOHOHFWLRQVLQ)ODQGHUV%HOJLXP %UDP:DXWHUV'ULHV9HUOHWƒ -RKDQ$FNDHUWƒƒ * University College Ghent and Ghent University Research Centre of the Flemish Government and University College Ghent Hasselt University Bram.Wauters@hogent.be Dries.Verlet@dar.vlaanderen.be Johan.Ackaert@uhasselt.be Paper prepared for ECPR General Conference, Potsdam 2009 Panel: New trends in electoral systems. Reforms towards more personalized electoral rules and elections. Section: Changes in Participation and Representation

,QWURGXFWLRQ Electoral systems are according to the classic definition of Rae (1967: 14) rules which govern the processes by which electoral preferences are articulated as votes and by which these votes are translated into distributions of governmental authority (typically parliamentary seats) among the competing political parties This translation from votes into seats is far from neutral. The choice for a particular electoral system can have farreaching consequences for the composition of a local council or a parliament. Three basic variables are important in this respect: ballot structure, district magnitude and electoral formula (Rae, 1967 ; Lijphart, 1994). The EDOORWVWUXFWXUH consists of three aspects (Van der Kolk, 2003): the number of votes a voter is allowed to cast, the type of information a voter can give (nominal, ordinal or numerical) and whether voters can vote for parties, for individuals or for both. The GLVWULFW PDJQLWXGH can be measured by dividing the number of seats by the number of districts (Taagepera & Shugart, 1989). The HOHFWRUDOIRUPXOD refers to the way votes are translated into seats: there are three main types: a majoritarian system with an absolute majority ('majority system'), a majoritarian system with a relative majority ('plurality system') en a system of proportional representation ( PR system ). Besides these main types, there are intermediary systems which combine elements of PR systems and majoritarian systems. We will focus in this article on the preferential vote, which is part of the ballot structure. The analysis will be restricted to PR list systems. Preferential voting, i.e. the possibility to vote for candidates, differs enormously between countries using PR (Van der Kolk, 2003). The type of electoral list, a crucial variable in this respect, refers to the main distinction between closed and open lists. Some PR systems allow voters to express their preferences for political parties only. Such systems are named closed list systems: the seats obtained by a party are automatically allocated to the candidates according to the list order as determined by the party. Other PR systems give voters the opportunity to change the list order put forward by political parties by allowing them to cast a 2

preferential vote for one or more candidates of the same party. In such an open list system, voters choose between individual candidates of a party. Seats are allocated according to the number of preferential votes obtained by the candidates. The voters decide who will be elected. In practice, however, many mixed systems exist: preferential votes play a partial role in determining who is elected, together with other considerations such as the list order (Karvonen, 2004). Some of these systems resemble more open-list systems, while others look more like closed-list systems. Whether they can be situated towards one pole or another is in some systems legally determined, while in other systems this depends upon the percentage of voters that cast a preferential vote. The latter is in particular the case for flexible lists (Marsh, 1985), where voters have the choice between supporting the list or voting for one or more preferred candidates. The candidates at the top of the list receive in such a system list votes in addition to their own preferential votes, which increases their chance to become elected. Where the percentage of preferential votes is low, the number of list votes to distribute is large and consequently, the list order is often decisive. In such a situation, the system resembles a closed-list system. In elections where the percentage of preferential voters is high, the number of list votes to distribute is limited, and as a consequence, the system will tend to be more like an open-list system (Wauters, 2000). The number of voters casting a preferential vote has thus a tremendous impact on who will be elected, but also on who is the main decision-maker in this process: parties or voters. In this way, this topic can be linked to recent discussions about a more personalised style of politics at the local level as a way of recovering the relation between citizens and politicians (Kersting, 2005). Due to a decline of the linkage function of (local) political parties, the gap between voters and local politicians has increased. It is believed that giving more weight to individual voters and individual politicians, at the expense of parties, could improve this relationship. Preferential votes play a crucial role in this tendency towards a more personalised style of politics. It could serve as a guarantee for a strong link between voters and politicians. Majoritarian systems produce, due to their small electoral districts, automatically a form of linkage between the elected and the electors. PR systems, which are by definition organised in larger districts, often 3

lack such a connection. A system of preferential voting in a PR system could, however, help to overcome this problem. Preferential votes do not only matter directly in determining who is elected, as was illustrated above, but also impacts indirectly on the composition of the local government. Research has revealed that for 80 % of the local parties in Flanders the number of preferential votes serves as an important criterion to designate executive positions in local government (Ackaert, 1996). A candidate who obtains a large number of votes has a good chance to become mayor or alderman. Consequently, in municipalities where many voters cast a preferential vote, they could also have an indirect, but substantial impact on the composition of the local government. Preferential voting also plays an important role in maintaining and fostering crucial values for the democratic functioning of a political system. A comparative analysis has shown that systems with preferential voting promote a greater sense of fairness about election outcomes among voters than systems without preferential voting. This sense of fairness is an important component of the citizen satisfaction with the democratic system (Farrell & McAllister, 2006). Despite the importance of preferential voting for the democratic functioning of (local) authorities, research attention for this topic has been limited. Preferential voting is often (but only briefly) mentioned in the literature as part of the broader electoral system, but the number of studies on preferential voting as a separate topic are scarce (Katz, 1986 ; Karvonen, 2004). It is revealing in this respect to note that different terms 1 are used to denote this phenomenon: preferential voting, personal voting, person voting, preference vote and intraparty choice (Karvonen, 2004). This probably also hampered (comparative) research on this topic. Moreover, most of the time, attention is mainly focused upon the possible effects of a system of preferential voting on variables such as government stability, factionalism, campaign financing, legislative turn-over and the presence of underrepresented groups 1 This phenomenon should, however, not be confused with the personal vote, which is a term reserved for the study of candidates in single-member constituencies. 4

(Karvonen, 2004 ; Matland, 2005). Other approaches aim to define and classify systems of preferential voting in comparative perspective (Marsh, 1985 ; Van der Kolk, 2007). In this article, we will study the factors stimulating the use of preferential votes. Research on the effect of factors influencing casting a preferential vote is underdeveloped (Van der Kolk, 2003). Since the exact meaning of preferential voting differs considerably accros countries, comparative research may not be very helpful in studying the causes for casting a prefential vote (Van der Kolk, 2003). Therefore, our analysis is an in-depth study of preferential voting in local elections in one single region (Flanders, a region of Belgium). This allows us to take a large number of explanatory variables into account. More in particular, we will analyse in what kind of Flemish municipalities voters are more likely to cast preferential votes. Before setting out the research hypotheses, we will describe the local electoral system in Flanders (Belgium). 3UHIHUHQWLDOYRWHVDWWKHORFDOOHYHOLQ%HOJLXP A Belgian voter has the choice between casting a preferential vote for one or more candidates (on a single party list) and casting a list vote. This kind of list system can be named a flexible list system (Marsh, 1985) or a weak preferential voting in list system (Karvonen, 2004). Candidates with a number of preferential votes attaining the threshold of eligibility are generally speaking automatically elected. The other candidates can make use of the list votes in order to reach the threshold. These list votes are distributed to the candidates according to the list order. As a consequence, the candidates at the top of the list have a substantial advantage, as a result of which the system for a long time functioned as a de facto closed list system, especially at the national level. For local elections, where the social distance between voters and elected is small, the percentage of voters casting a preferential vote has always been high, resulting in a relatively high number of candidates elected out of the list order (Wauters, 2000 ; Ackaert et al, 2007). The local electoral system has always resembled more an open-list system. 5

In order to enhance the impact of preferential votes in national and regional elections, a recent electoral reform halved the impact of the list votes on the allocation of the seats, as a result of which the advantage of higher ranked candidates was weakened (Wauters, 2003). Belgium has thus evolved in practice from a semi-closed list PR system to a semiopen list PR system. Also for elections at the local level in Flanders, where the impact of preferential votes was nevertheless already high, new electoral laws halving the transfer of list votes and eventually taking only one third of the list votes 2 were adopted (Decree of 7 July 2006 concerning adaptations on the Local Electoral Decree). Local elections are held every six years. The most recent local elections were held in 2006, 2000 and 1994. The power of the preferential vote has undergone changes over these elections: in 1994, the total number of list votes was transferred according to the list order, while in 2006 still only one third of these votes were used. The power of a preferential vote, and hence the incentive to cast one, has thus increased. It is one of the aims of the article to investigate the impact of these institutional changes on voting behavior (whether or not casting a preferential vote). 5HVHDUFKTXHVWLRQV In this paper, an analysis on the macro-level will be conducted: characteristics of local communities will be linked to preferential voting on the level of a local community. Our analysis will be twofold. We will first look at the situation in 2006 and the variables that could explain the share of preferential votes. Our dependent variable here is the percentage of the voters that have cast a preferential vote, calculated on the total number of voters in a municipality. Secondly, the impact of the change in electoral laws will be evaluated by looking at the difference in the share of preferential votes between 1994 and 2006. The difference between the two percentages (1994 and 2006) will serve as dependent variable. Five groups of independent variables will be brought into the analysis: 2 In fact, the number of list votes per party is multiplied by the number of seats per party (as was done before), and then divided by three. 6

6RFLRGHPRJUDSKLFYDULDEOHV: age, social class, gender and ethnic origin Marsh (1985) approaches preferential voting with instruments to analyse participation behaviour. The resource theory states that participation depends upon the resources one has (Verba, Nie & Kim, 1978). Resources include among others money, education, time, civic engagement and access to political information. Some members are better equipped than others to participate. From several studies, it appears that in general women, young (or conversely rather old) people, non-white people and people from lower socioeconomic classes often lack these vital resources, and as a consequence participate less in political activities. Marsh states that these variables also influence negatively the chance to cast a preferential vote. On the contrary, we should look also at research that has shown that ethnic minority candidates and candidates of lower social classes have a higher chance to become elected in areas where their social group has a large concentration in the population (Anwar, 2001). Preferential voting for candidates of these underpriveleged groups could be seen as a strategy to enhance their representation. Consequently, preferential voting could be more extensive in communities where members of such underprivileged groups are numerous, which is the opposite expectation of the resource theory. Moreover, foreigners were (within certain conditions) for the 2006 local election for the first time allowed to vote. They had, unlike Belgian voters, to register themselves as voter. It is not unthinkable that these citizens demonstrate by their personal registration a larger degree of political commitment and develop a clearer view on politics and politicians and that by consequence they incline more to use the preferential vote. Variables typically associated with particaption levels, such as age, social status, gender and ethnic origin, will be included in the analysis. Earlier research has already investigated the effect of some of these variables on the individual level. As for age, Van der Kolk (2003) found that younger and older people cast less a preferential vote than people between 30 and 50 years old. The evidence about the effect of gender on preferential voting yields, however, a mixed picture: in Denmark and Norway, men use preferential voting more than women, while in the Netherlands, women use preferential 7

voting more often (Van der Kolk, 2003). We should note that contrary to the studies described above, the effect of these variables will be tested here on the macro level. 3ROLWLFDOYDULDEOHV: number of parties, ideology of parties and presence of local parties As set out earlier, preferential voting offers an extra choice above the choice a voter has between parties. We hypothesize here that when the choice between parties is limited in a local community, because only a few parties are putting forward candidate lists, voters will make more extensively use of preferential voting. By making a choice between candidates, a voter can still make his preferences known despite the limited choice between parties. Also the effect of the ideology of the major parties in a local community will be analysed. In general, preferential voting seems to be more common amongst right-wing parties than amongst left-wing parties (Hessing, 1985 ; Wauters & Weekers, 2008), but evidence differs here from one country to another (Marsh, 1985). It can also be hypothesized that rather new parties and parties that are mainly waging a party campaign instead of individual campaigns (mainly green and extreme right parties) (Weekers & Maddens, 2009), are more likely to attract fewer preferential votes, and consequently that municipalities where these parties perform well in elections will exhibit lower percentages of preferential votes. A final aspect of the political variables is constituted by the presence of local lists. These local lists represent a broad range of parties not operating under the heading of a national party at the local elections. Despite their diversity, they share some characteristics: local parties are in general less structured, less ideologically outspoken, more oriented towards municipal issues and give local notables (often known from the local associational life) a more prominent position on their candidate lists (Steyvers et al, 2008). Consequently, it could be expected that local lists attract more preferential votes, and that municipalities with a large share of local lists will have a higher percentage of preferential voting. 6RFLDOFDSLWDOYDULDEOHV: number of associations Two components of social capital can be distinguished: structural and cultural components (Putnam, 2000). One structural element of social capital that is relevant here is the associational 8

life of a community. Voluntary organizations based on membership can be seen as schools of democracy where social and civic skills are learnt. Moreover, voluntary organizations play an important role in the recruitment of candidates. We hypothesize that a flourishing associational life increases the chance that people are aware of the value of a preferential vote (political education) and that candidates tend to be better known to a large share of the electorate. The associational life of a community is measured by the number of voluntary associations at the local level as proxy. The choice for this indicator is based on practical and substantive reasons. Recent data on the number of associations at the local level in Flanders are available (Lauwereysen & Colpaert, 2004). Moreover, this proxy serves as a particularly good indicator for social capital as the formal voluntary association is the most important form of horizontal interaction and reciprocity. *HRJUDILFYDULDEOHV: population density and number of inhabitants Studies contradict each other as to whether an urban or a rural environment is beneficial for attracting a large share of preferential voters (Marsh, 1985). For Belgium, however, it has been shown that for the local elections, there is a straightforward relationship between the population density and the percentage of preferential votes of a municipality: the more urban a municipality, the less preferential votes (Wauters, 200). This relationship was explained by the concept of social distance : in rural communities politicians are more close to the population than in large cities, and hence people are more likely to cast a preferential vote for these locally well-known candidates. In a second step, the effect of the electoral reforms giving more weight to preferential votes will be investigated: have they led to an increase in the use of preferential votes, and if so, in what kind of local communities more than in other? The increase or decrease in the percentages of voters casting a preferential vote is here the dependent variable. The investigation over time of the effect of the electoral reform can be linked to the influence of LQVWLWXWLRQDO YDULDEOHV, a fifth group of explanatory variables. The most obvious incentive to cast a preferential vote is, in Marsh (1985) view, the expectation that this kind of vote will affect the distribution of seats. It seems logic that when the impact of list votes is diminished by electoral reforms and as a consequence the weight given to preferential votes grows, the percentages of voters casting a preferential vote 9

will increase. It will be investigated whether this indeed occurs by comparing the percentages of preferential votes between the local elections of 1994, where the total number of list voters is transferred to candidates according to the list order and those of 2006, where only one third of the list votes is transferred. 5HVXOWV %LYDULDWHDQDO\VHVIRUWKHHOHFWLRQV We will start by investigating the bivariate relationships between the percentage of preferential votes and a large number of variables that can be catalogued under one of the groups of variables discussed in the previous section. As for the socio-demographic variables, our results show that municipalities with a rather old and female population tend to exhibit a lower percentage of preferential votes. This is line with the resource theory. The share of foreigners and the share of young people in the population does not have a significant effect, however. The analysis on socioeconomic variables yields a mixed picture: the share of social housing in a municipality correlates with the percentage preferential votes, while the long-term employment rate does not (not in table). 7DEOH%LYDULDWHUHODWLRQVKLSVEHWZHHQYDULDEOHVDQGWKHSHUFHQWDJHRI SUHIHUHQWLDOYRWHVLQRQWKHOHYHORIWKHPXQLFLSDOLW\1 R P 6RFLRGHPRJUDSKLFYDULDEOHV Percentage 18-64 years old.121* 0.033 Percentage younger than 35 years old.050 0.379 Percentage older than 65 years old -.150** 0.009 Percentage older than 80 years old -.151** 0.008 Percentage women -.419** 0.000 Percentage foreigners.081 0.157 Share of social housing -.140* 0.014 3ROLWLFDOYDULDEOHV Number of parties in elections -.444** 0.000 10

Percentage of CD&V-N-VA.056 0.327 Percentage of Sp.a-Spirit -.133* 0.020 Percentage of OpenVLD -.139* 0.015 Percentage of Vlaams Belang -.499** 0.000 Percentage of Groen! -.192** 0.001 Vote percentage for local lists.253** 0.000 6RFLDOFDSLWDOYDULDEOHV Number of associations per 1000 inhabit..392** 0.000 *HRJUDSKLFYDULDEOHV Number of inhabitants -.281** 0.000 Population density -.453** 0.000 * p <.05 ; ** p <.01 Secondly, almost all political variables significantly correlate with the percentage of preferential votes. Both the number of parties and the type of parties are related to preferential voting as measured on the local level. As for this second element, the vote shares of the extreme right Vlaams Belang and the ecologist party Groen! shows, as expected, the strongest (negative) correlation with the percentage of preferential votes. As for the vote share of local lists, there also is a significant correlation. Thirdly, also the number of associations, as proxy for social capital, correlates significantly with the percentage of preferential votes. This confirms the social capital hypothesis. Finally, all demographic variables included in the analysis have a significant negative relationship with the use of preferential voting: the more urban a municipality, the less preferential votes are cast. 0XOWLYDULDWHDQDO\VLV As it seems that many variables from the explanatory framework are correlated with each other, it is advisable to run a multivariate analysis. 7DEOH0XOWLYDULDWHUHJUHVVLRQPRGHOZLWKWKHSHUFHQWDJHRISUHIHUHQWLDOYRWHVLQ RQWKHOHYHORIWKHPXQLFLSDOLW\DVGHSHQGHQWYDULDEOH 11

B Std. error Beta P &RQVWDQW 111.424 28.275.000 6RFLRGHPRJUDSKLFYDULDEOHV Percentage 18-64 years old.288.155.104.063 Percentage women -80.878 42.613 -.122.059 Percentage foreigners 23.593 5.522.206.000 Share of social housing -10.244 7.839 -.063.192 3ROLWLFDOYDULDEOHV Number of parties in elections -.448.237 -.134.060 Percentage of CD&V-N-VA -.008.060 -.029.889 Percentage of Sp.a-Spirit -.003.061 -.007.961 Percentage of OpenVLD -.028.062 -.070.652 Percentage of Vlaams Belang -.162.067 -.248.015 Percentage of Groen! -.064.079 -.054.413 Percentage of local parties -.004.059 -.022.946 6RFLDOFDSLWDOYDULDEOHV Number of associations per 1000 inhabit..952.268.200.000 *HRJUDSKLFYDULDEOHV Number of inhabitants.000.000 -.022.703 Population density -.002.001 -.155.012 $GM5ð S1 This multivariate analysis shows that four variables significantly affect the percentage of preferential votes in a municipality (one from each category of variables: the percentage of foreigners, the vote percentage of the extreme right Vlaams Belang, the number of associations and the population density). The latter three are in line with the bivariate analyses, but the strong significant effect of the share of foreigners in the population comes as a surprise, since there was no significant effect in the bivariate analysis. We do not have a straightforward explanation for this phenomenon, except maybe a sort of compensation effect for urban areas. We should note that the effects of the percentage of women, the percentage of 18-64 years old and the number of parties are only slightly non-significant at 0.05-level. 12

7KHHIIHFWRIWKHHOHFWRUDOUHIRUPVELYDULDWHDQDO\VHV Here, the impact of the electoral reforms will be assessed. Between 1994 and 2006, electoral reforms enhancing the impact of preferential voting were adopted (first halving and eventually taking only one third of the list votes for the distribution of seats). 7DEOH7KHSHUFHQWDJHRISUHIHUHQWLDOYRWHVDQGWKHDYHUDJHQXPEHURI SUHIHUHQWLDOYRWHVLQDQGPHDQVDQGVWDQGDUGGHYLDWLRQ <HDU 1 0HDQ 6WDQGDUG GHYLDWLRQ Percentage 1994 308 84.36 7.39 preferential votes 2006 307 85.63 5.30 As can be seen from Table 3 the general effect on the average percentage of preferential votes has not been tremendous: only a slight increase (from 84,36 % to 85,63 %) could be noted. This is confirmed by a correlation coefficient of 0.597 between the percentages of preferential votes on the municipal level in 1994 and 2006 (not in the table). The rather marked drop in the standard deviation between 1994 and 2006, however, seems to suggest that the electoral reforms have rendered the preferential vote percentages across municipalities more equal. It seems plausible to assume that in municipalities with low percentages of preferential votes these percentages have increased between 1994 and 2006, while in municipalities with high percentages they have remained stable or have witnessed a slight decrease. Whether this is indeed so will be investigated in the subsequent analysis. 7DEOH%LYDULDWHUHODWLRQVKLSVEHWZHHQYDULDEOHVDQGWKHGLIIHUHQFHLQWKH SHUFHQWDJHRISUHIHUHQWLDOYRWHVEHWZHHQDQGRQWKHOHYHORIWKH PXQLFLSDOLW\1 R P 6RFLRGHPRJUDSKLFYDULDEOHV Percentage 18-64 years old -.028 0.619 Percentage younger than 35 years old.203** 0.000 Percentage older than 65 years old -.154** 0.007 Percentage older than 80 years old -.120* 0.036 Percentage women.144* 0.011 Percentage foreigners.384** 0.000 13

Share of social housing.191** 0.001 3ROLWLFDOYDULDEOHV Number of parties in elections.228** 0.000 Percentage of CD&V-N-VA -.075 0.189 Percentage of Sp.a-Spirit -.038 0.506 Percentage of OpenVLD -.048 0.404 Percentage of Vlaams Belang.290** 0.000 Percentage of Groen!.189** 0.001 Number of local lists in elections -.028 0.625 6RFLDOFDSLWDOYDULDEOHV Number of associations per 1000 inhabit. -.238** 0.000 *HRJUDSKLFYDULDEOHV Number of inhabitants.238** 0,000 Population density.369** 0.000 * p <.05 ; ** p <.01 The bivariate analyses in Table 4 (comparison 1994-2006) show in general the opposite picture of Table 1 (2006). This implies that in municipalities where the percentage of preferential votes is already high, the effect of the electoral reform is limited. The largest increase could be noted in localities where the percentage of preferential votes was rather low in 1994. There is, for instance, a negative correlation between number of associations and the evolution in preferential voting. This means that in municipalities where the number of associations is high, the percentage of preferential votes is already high (as was shown in Table 1) and there was no room for a further large increase. Conversely, in municipalities with only a few number of associations, the percentage of preferential voting is low, and hence the effect of the electoral reform is rather large. An exception to this reverse pattern is formed by the share of older people in the population: this correlates negatively with both the percentage as such as with the evolution of the percentage of preferential votes. Perhaps, older people are less informed about these reforms or are less familiar with the practice of (electronic) preferential voting. Also the percentage of foreigners in the population correlates positively with both the percentage in 2006 and the evolution of this percentage (albeit the former correlation is not significant). 14

7KHHIIHFWRIWKHHOHFWRUDOUHIRUPVPXOWLYDULDWHDQDO\VHV 7DEOH0XOWLYDULDWHUHJUHVVLRQPRGHOZLWKWKHGLIIHUHQFHLQWKHSHUFHQWDJHRI SUHIHUHQWLDOYRWHVEHWZHHQDQGRQWKHOHYHORIWKHPXQLFLSDOLW\DV GHSHQGHQWYDULDEOH1 B Std. error Beta P &RQVWDQW 55.003 35.744.125 6RFLRGHPRJUDSKLFYDULDEOHV Percentage 18-64 years old -.203.195 -.065.299 Percentage women -85.246 53.871 -.114.115 Percentage foreigners 39.411 6.981.306.000 Share of social housing 8.204 9.910.045.408 3ROLWLFDOYDULDEOHV Number of parties in elections -.011.300 -.003.971 Percentage of CD&V-N-VA -.018.076 -.056.810 Percentage of Sp.a-Spirit -.071.077 -.140.359 Percentage of OpenVLD -.030.078 -.067.699 Percentage of Vlaams Belang.094.084.127.264 Percentage of Groen!.201.099.149.044 Percentage of local parties -.023.074 -.112.756 6RFLDOFDSLWDOYDULDEOHV Number of associations per 1000 inhabit. -.214.339 -.040.528 *HRJUDSKLFYDULDEOHV Number of inhabitants.000.000.042.519 Population density.003.001.234.001 $GM5ð S1 Here again the percentage of foreigners in a local population significantly affects the increase in the percentage of people casting a preferential vote and once again, we do not have a direct straightforward explanation. A factor that could be relevant in this respect is that foreigners were granted in 2006 for the first time the right to vote in elections in Belgium. Perhaps, these new voters, less familiar with Belgian parties, have casted a preferential vote. Also the fact that foreigners had to register, and consequently mainly motivated and well-informed voters could vote, could be an explanation for this. 15

The population density also impacts on our dependent variable: the more urban a municipality, the larger the effect on preferential voting of the electoral reforms. It seems that the electoral reforms has had most effect in cities, while they were rather superfluous in more rural municipalities because the share of preferential voting was already high there. The vote shares of two smaller, new parties (extreme right Vlaams Belang and the ecologist Groen!) have an influence on the percentages of preferential voting. This effect is probably not only due to the electoral reform, but could also be well explained by the increasing professionalization and personalization of these parties that become more and more established. Also a similar reasoning as above could be made: these are parties with a very low share of preferential votes, and consequently, there was more room for an increase than among other more traditional parties. &RQFOXVLRQ Preferential voting has not yet extensively investigated as a separate research object. This phenomenon is important, however, since it determines for instance whether a flexible list system, as is used in Belgium, functions rather as an open list-system or as a closedlist system. It could also be related to the efforts to introduce a more personalized style of politics. In this paper, we have investigated in what kind of municipalities voters are more likely to cast a preferential vote and whether the electoral reform granting the voters more power had an effect in what kind of municipalities. We have put forward four groups of variables that could explain both the actual percentage of preferential votes at the municipal level and the evolution of this percentage as influenced by the electoral reforms. These groups were socio-demographic variables, political variables, social capital variables and geographic variables. Our analysis shows that variables from each group correlate significantly with the percentage of preferential votes (even in a multivariate analysis). The comparison between 1994 and 2006 often yields the reverse picture: characteristics of municipalities that have a positive 16

effect on the percentage of preferential votes, have a negative impact upon the evolution between 1994 and 2006, and vice versa. This results in the electoral reform having only an outspoken marked effect in urban municipalities, because elsewhere local politics is already to a large extent personalized by locally known politicians. In these rural municipalities the electoral reform was superfluous. 17

5HIHUHQFHV Ackaert, J. (1996), De kiezers delen de kaarten uit, de partijen spelen ermee, in: J. Buelens & K. Deschouwer (reds), 'H 'RUSVVWUDDW LV GH :HWVWUDDW QLHW, Brussel: VUBPress, pp. 49-74 Ackaert, J., De Ceuninck, K., Reynaert, H., Steyvers, K. & Valcke, T. (2007), Doen (wijzigende) instituties ertoe? De invloed van het gemeente(kies)decreet op de gemeenteraadsverkiezingen van 2006, in: 5HV3XEOLFD, 49(1), pp. 15-33 Anwar, M. (2001), The participation of ethnic minorities in British politics, in: -RXUQDORI (WKQLFDQG0LJUDWLRQ6WXGLHV, 27(3), pp. 533-549 Farrell, D.M. & McAllister, I. (2006), Voter satisfaction and electoral systems: Does preferential voting in candidate-centred systems make a difference?, in: (XURSHDQ -RXUQDORI3ROLWLFDO5HVHDUFK, 45(5), pp. 723-749 Hessing, R.C. (1985), Bij voorkeur: een onderzoek naar het gebruik van voorkeurstemmen, in: $FWD3ROLWLFD, (2), pp. 157-176. Karvonen, L. (2004), Preferential voting: Incidence and Effects, in:,qwhuqdwlrqdo 3ROLWLFDO6FLHQFH5HYLHZ, 25(2), pp. 203-226 Katz, R. (1986), Intraparty preference voting, in: B. Grofman and A. Lijphart (eds.), (OHFWRUDOODZVDQGWKHLUFRQVHTXHQFHV. New York: Agathon Press, pp. 85-103 Kersting, N. (2005), Reforming local electoral systems in Europe, in: H. Reynaert, K. Steyvers, P. Delwit & J-B Pilet (eds), 5HYROXWLRQRU5HQRYDWLRQ"5HIRUPLQJORFDOSROLWLFV LQ(XURSH, Brugge: Vanden Broele, pp. 29-48. Lauwereysen, K. & Colpaert, J. (2004). 'HORNDOHDIGHOLQJHQYDQGHVRFLRFXOWXUHOHRUJDQLVDWLHV LQNDDUWJHEUDFKWSocius gegevens 3, 53-71. Lijphart (1994) (OHFWRUDO6\VWHPVDQG3DUW\6\VWHPV$6WXG\RI7ZHQW\VHYHQ'HPRFUDFLHV, New York: Oxford University Press. Marsh, M. (1985), The voters decide? Preferential voting in European list systems, in: (XURSHDQ-RXUQDORI3ROLWLFDO5HVHDUFK, 13(4), pp. 365-378 Matland, R.E. (2005), Enhancing women s political participation: legislative recruitment and electoral systems, in: J. Ballington & A. Karam (ed.), :RPHQLQ3DUOLDPHQWEH\RQG QXPEHUV, Stockholm: IDEA, pp. 93-111. Putnam, R. (2000). %RZOLQJDORQH7KHFROODSVHDQGUHYLYDORI$PHULFDQFRPPXQLW\New York: Simon & Schuster. 18

Rae, D. (1967), 7KH SROLWLFDO FRQVHTXHQFHV RI HOHFWRUDO ODZV. New Haven: Yale University Press. Steyvers, K., Reynaert, H., De Ceuninck, K. & Valcke, T. (2008), All politics is local, partisan or national? Local lists in Belgium, in: M. Reiser & E. Holtmann (eds.), )DUHZHOO WR WKH 3DUW\ 0RGHO",QGHSHQGHQW /RFDO /LVWV LQ (DVW DQG :HVW (XURSHDQ &RXQWULHV, Wiesbaden: VS Verlag, pp. 169-194, Taagepera, R. & Shugart, M.S. (1989), 6HDWVDQGYRWHV7KHHIIHFWVDQGGHWHUPLQDQWVRI HOHFWRUDOV\VWHPV, New Haven: Yale University Press. Van der Kolk, H. (2003), Preferential voting in European local elections. Opportunities, use, justifications and explanations in Denmark, Germany, the Netherlands and Norway, Paper presented at the ECPR General Conference, Marburg, 18-21 September 2003. Van der Kolk, H. (2007), Local electoral systems in Europe, in: /RFDO *RYHUQPHQW 6WXGLHV, 33(2), pp. 159-180 Verba, S., Hie, N. H. & Kim, J.-O. (1978), 3DUWLFLSDWLRQDQG3ROLWLFDO(TXDOLW\$6HYHQ QDWLRQ&RPSDULVRQ. Cambridge: Cambridge University Press. Wauters, B. (2000), 'HNUDFKWYDQGHYRRUNHXUVWHPLQVWDGHQGRUS(HQDQDO\VHELMGH JHPHHQWHUDDGVYHUNLH]LQJHQ YDQ LQ 9ODDQGHUHQ, Leuven: Afdeling Politologie KU Leuven. Wauters, B. (2003), Het gebruik van voorkeurstemmen bij de federale parlementsverkiezingen van 18 mei 2003, in: 5HV3XEOLFD, 45(2-3), pp. 401-428 Wauters, B. & Weekers, K. (2008), Het gebruik van de voorkeurstem bij de federale parlementsverkiezingen van 10 juni 2007, in: 5HV3XEOLFD, 50(2), pp. 49-88 Weekers, K. & Maddens, B. (2009), +HWJHOGYDQGHSDUWLMHQ, Leuven: Acco. 19

$SSHQGL[ Descriptive statistics of the variables used in this analysis 9DULDEOHQDPH 1 0HDQ 6WDQGDUGGHYLDWLRQ Percentage preferential votes 2006 307 85,63 5,30 Percentage preferential votes 1994 308 84,36 7,39 6RFLRGHPRJUDSKLFYDULDEOHV Percentage 18-64 years old 307 62.30 1.92 Percentage younger than 35 years old 307 40.27 2.35 Percentage older than 65 years old 307 17.52 2.21 Percentage older than 80 years old 307 4.22 0.89 Percentage women 307 50.46 0.80 Percentage foreigners 307 3.47 4.64 Share of social housing 307 4.02 3.25 Share of long-term unemployment 307 1.31 0.51 3ROLWLFDOYDULDEOHV Number of parties in elections 308 4.76 1.58 Number of parties in the local council 308 4.15 1.07 Percentage of CD&V-N-VA 3 308 30.63 18.51 Percentage of Sp.a-Spirit 308 9.72 11.77 Percentage of OpenVLD 308 13.87 13.30 Percentage of Vlaams Belang 308 10.55 8.07 Percentage of Groen! 308 2.63 4.43 Vote percentage for local lists 308 31.59 28.91 Number of national parties in elections 308 3.24 1.57 Number of local lists in elections 308 1.52 1.13 6RFLDOFDSLWDOYDULDEOHV Number of associations per 1000 inhabit. 307 2,62 1,11 *HRJUDSKLFYDULDEOHV Number of registered voters 308 15091.30 22746.39 Number of councilors 308 23.89 6.25 Number of inhabitants 307 19696.08 31524.30 Population density 307 521.30 448.70 3 The means of the percentages of the parties are calculated on the basis of the percentages of these parties by municipality. Consequently, these means do not correspond to the overall score of parties as calculated over the whole region. 20